The Voynich Ninja

One should never underestimate the imagination of a person.
Here is an example:
You are not allowed to view links. Register or Login to view.

This is a simple one-to-one encryption. Even NSA couldn't crack it.
The VM shows a similar behavior. It is easy to say that it is not a natural language just because you don't understand it.
But that is exactly what makes a good cryptogram.
I've been working on VM for a while now, and yes it's a natural language.

All conclusions based on detailed statistics or general (visual) inspections are based on a set of assumptions.

Unfortunately, these assumptions are rarely stated, and often not even fully realised.

The tacit assumption in almost all analyses is that one Voynich character represents one plain text character or one sound.

The tentative conclusion that Voynichese is incompatible with human language is based on this assumption.

Statistical tools are an excellent way to test the assumptions, but this is rarely done.

As far as I am concerned, Voynichese is not incompatible with human language in general. This has not been demonstrated by any test.

Also under the assumption that one symbols is one PT character or sound, Voynichese has not been demonstrated to be incompatible with human language.
It behaves rather similarly to oriental languages that exhibit a clear syllable structure.

Now, to be clear, I am not saying that this is what Voynichese is. I am just saying that it is not incompatible with that.

(10-03-2020, 08:51 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view....
The tacit assumption in almost all analyses is that one Voynich character represents one plain text character or one sound.

Yes, I agree. Most of the claimed "solutions" and the majority of computational attacks make this assumption. Unfortunately, this assumption is completely at odds with how most text was written in the later medieval period, and is also at odds with the way some ciphers were designed.

Quote:The tentative conclusion that Voynichese is incompatible with human language is based on this assumption.

I'm not sure this is always true. Many of those who make this assumption come up with "solutions" that have been somewhat "fixed" with subjective adjustments and do think the VMS is natural language.

Quote:Statistical tools are an excellent way to test the assumptions, but this is rarely done.

Agreed. Many of the statistical studies go on the one-to-one assumption of some underlying language without starting at Square 1 to see WHAT it actually is and how it is actually structured.

Quote:As far as I am concerned, Voynichese is not incompatible with human language in general. This has not been demonstrated by any test.

As it stands, I believe it is incompatible with natural language without some significant processing, and I believe I can demonstrate this, but I am having difficulty finding enough time to describe it so people can understand it. I have a paper on it, but it requires a lot of examples to show step-by-step why this is so.

Quote:Also under the assumption that one symbols is one PT character or sound, Voynichese has not been demonstrated to be incompatible with human language.
It behaves rather similarly to oriental languages that exhibit a clear syllable structure.

Again, I believe it is incompatible with natural language without some significant processing, and I believe I can demonstrate this. As far as oriental languages go, I am not fluent in them, but I know some Japanese and Korean (and very tiny bit of Chinese). At one point I could read news articles in Korean and understand simple conversation (I am quickly losing these skills due to lack of use) and I feel very strongly that as it stands (without processing) that Voynichese is more similar to these than to western languages IN SOME WAYS but... and this is a huge "but", Voychinese completely fails to exhibit the syllabic positional variety that is characteristic of Asian languages.


I think increased study of abjadic characteristics or numerical possibilities (or the possible interpretation of shape variations as different characters) might be more productive than "solutions" that assume natural language from the get-go without actually examining the characteristics of the text.

A "black swan" I'd like to see is a long text with an average 3% rate of Reduplication+QuasiReduplication. If such a bird exists, it likely is some Eastern language. It could be written in a non-alphabetic script, so it is not easily spotted by a character-by-character computational analysis. For instance, Rene pointed out that Thai has a special "reduplication-character" which is used instead of actual repeated words: in order to compare such a script with the hypothesis that Voynichese is phonetic, one should first convert the Eastern script into a phonetic system (e.g. IPA).

It becomes really interesting when a German speaker needs an interpreter for German.
We in Zurich speak a lot with ( cha, che, cho, chi, chu ).
In the example the alphabet does not even contain ( C ). But it is the same language.
How do I want to judge something if I decipher it, if I don't understand it, even if I am right.
Now I am still in the 1400 century.
Who understands it, even if it is not encrypted.

You are not allowed to view links. Register or Login to view.

@Helmut Winkler,
It falls into your dialect area (Bavarian). Do you understand it ? I stand at my limits.

[quote="Aga Tentakulus" pid='35322' dateline='1583837654']
It becomes really interesting when a German speaker needs an interpreter for German.
We in Zurich speak a lot with ( cha, che, cho, chi, chu ).
In the example the alphabet does not even contain ( C ). But it is the same language.
How do I want to judge something if I decipher it, if I don't understand it, even if I am right.
Now I am still in the 1400 century.
Who understands it, even if it is not encrypted.

You are not allowed to view links. Register or Login to view.

@Helmut Winkler,
It falls into your dialect area (Bavarian). Do you understand it ? I stand at my limits.
[/quot

I am Bavarian, but my own dialect is Franconian, which is Mitteldeutsch, not Oberdeutsch  as Bavarian and the dialect from the Zillertal. I understand most of it, some of the vocabulary is strange, it is like Austrian, which has the same grammar as Hochdeutsch, but some different vocabulary, but there is nothing I don't understand, Karfiol is Blumenkohl and Paradeiser Tomaten. I have much more difficulties with Alemannic and North German dialects. Just to mention it, when I was in England, I had not much difficulties with the North countty dialects, which the Southern Brits had difficulties to understand. I think it depends much on how the single areas are related and in contact with one another Bavarians usually have more contact with Austrians than with North Germans

It is very hard to answer this question as a simple yes/no question. But I do think that you bring out some important points in your post, and then without going to deep dive into a lot of statistical "mumbo-jumbo", it would maybe help a little to bring out the, at least for me, most simple and outstanding feature of the Voynich script, based on a very simple metric. That is: the lack of longer repeated sequences (if you first of all assume a phonetic writing system). I have yet not found any text written by humans and containing meaningful content which lacks longer repeated sequences of 4-5 words or more. There always appears to be some redundancy in human language to a given degree, for a text as long as the Voynich text. But this itself does not automatically lead one to automatically believe that the Voynich manuscript contains just meaningless text. 

First, the text could be mostly meaningless, containing some meaning here and there. Perhaps due to being written in a very confused or obfuscated way, either deliberate or not. We do not know what type of document it is or it's purpose. It is also very interesting that analyzing a badly degraded meaningful text with lots of uncertainties in its character-interpretations will bring out similar statistical repetition patterns as the Voynich-transliterated text, i.e. many repeated 2-word or 3-word sequences, but basically absence of longer sequences. But the text is very clearly written and not badly degraded, and it does not seem to have a lot of ambiguities in its character-glyph encoding, so what is the explanation?

Second, the most brought out discussion, that it has been encrypted. But this has to then be in a way that makes it retain a certain structure and non-randomness, which is unusual or very hard to accomplish for longer texts.

Apart from the lack if longer sequences, you could also make the claim that the frequent repetitions of the same word one after another is incompatible with natural texts. The Voynich manuscript contains about twice as many consequent word-repetitions as would be expected of a randomly word-shuffled text. Indeed, most natural language writings contain many fewer consequent repetitions than their random counterpart. So that basically while identical words in most natural texts tends to avoid each other in proximity, the Voynich words behave more like magnets and wants to stick together. It is therefore also interesting to note that there is in fact one well known text in natural language containing meaning that has words behaving like magnets instead of what is usual for a natural writing, and it also shows about twice as many repetitions as its random counterpart: the Finnish work "Kalevala", collated by Elias Lönnrot (1835) (available from You are not allowed to view links. Register or Login to view.). This is an example of a poetic writing/prose and this particular found "black swan" perfectly shows the tendency to perhaps sometimes get too misled by our own cultural expectations and assumptions of writings, given the "typical" (and most statistical analyses will be based on the typical samples due to the lack of availability of a few good atypical candidates) - and we know that the Voynich book is a very atypical book, so why should we expect anything else of the text it contains?

(10-03-2020, 07:51 PM)Alin_J Wrote: You are not allowed to view links. Register or Login to view.Apart from the lack if longer sequences, you could also make the claim that the frequent repetitions of the same word one after another is incompatible with natural texts. The Voynich manuscript contains about twice as many consequent word-repetitions as would be expected of a randomly word-shuffled text. Indeed, most natural language writings contain many fewer consequent repetitions than their random counterpart. So that basically while identical words in most natural texts tends to avoid each other in proximity, the Voynich words behave more like magnets and wants to stick together. It is therefore also interesting to note that there is in fact one well known text in natural language containing meaning that has words behaving like magnets instead of what is usual for a natural writing, and it also shows about twice as many repetitions as its random counterpart: the Finnish work "Kalevala", collated by Elias Lönnrot (1835) (available from You are not allowed to view links. Register or Login to view.). This is an example of a poetic writing/prose and this particular found "black swan" perfectly shows the tendency to perhaps sometimes get too misled by our own cultural expectations and assumptions of writings, given the "typical" (and most statistical analyses will be based on the typical samples due to the lack of availability of a few good atypical candidates) - and we know that the Voynich book is a very atypical book, so why should we expect anything else of the text it contains?

Thank you, Jonas, that sounds interesting!
I am not sure I understand correctly the kind of repetition in Kalevala. Is it perfect-reduplication like e.g. daiin daiin? If so, is it anywhere close to the almost 1% rate in the VMS?

A while ago, Koen pointed out Latin poetry with several occurrences of immediately repeated words (his "MiscCarmina" collection, discussed You are not allowed to view links. Register or Login to view.). But in that case the perfect-reduplication rate was still too low (about 0.5%). Moreover, quasi reduplication is extremely rare in that case (about 0.1%).

The lack of longer repeated sequences is also an example of a statistic based on certain assumptions:
- the word groups we see are indeed words
- spelling is reasonably consistent
- there are no null characters
- the different character shapes are indeed different letters

The last bullet may be surprising, and it is one of these 'automatic' assumptions. However, if one just made the though experiment that e , ee , eee are equivalent, then the number or repeated sequences will increase.

Exact word repetitions are a bit more tricky. Of the four bullets above, only the first one would still remain valid.

(11-03-2020, 12:01 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.Thank you, Jonas, that sounds interesting!
I am not sure I understand correctly the kind of repetition in Kalevala. Is it perfect-reduplication like e.g. daiin daiin? If so, is it anywhere close to the almost 1% rate in the VMS?

I have only seen translations of this text, but IIRC, the repetitions are patterns like this:
The men go to the meadows,
the men go to the pastures,
the men go to the fields.
They sing songs of their country,
they sing songs of their families,
the sing songs of freedom...

and so on ... and on ... and on.

Fun fact, Borges named his most famous detective "Lönnrot". He didn't end up well.